DISCRETE MATHEMATICS ELSEVIER
Discrete Mathematics 177 (1997) 223-243
The weight hierarchies and generalized weight spectra of the projective codes from degenerate quadrics Zhexian Wan, Xinwen Wu* Institute of Systems Science, Chinese Academy of Sciences, Belting 100080, China Received 9 August 1995; revised 4 June 1996
Abstract
The weight hierarchies and generalized weight spectra of the projective codes from degenerate quadrics in projective spaces over finite fields are determined. These codes satisfy also the chain conditions.
I. Introduction
Motivated by applications in cryptography, Wei [I I] introduced a generalization of the minimum Hamming weight of a linear code. Let O:q be a finite field, where q is a prime power. For any code D of block length n over O:q, define the support ~(D) by
x(D) = {i l ci ~ 0 for some (cl,c2 ..... cn)ED}, and the support weight ws(D) by
ws(D)
=
Iz(D)I.
Now let C be a linear [n, k] code over 0:q. For any r, where 1 ~
{dl(C),d2(C) ..... dk(C)}. * Corresponding author. E-mail:
[email protected]. Present address: Institute of Mathematics, Chinese Academy of Sciences, Beijing 100080, China. 0012-365X/97/$17.00 Copyright (~ 1997 Elsevier Science B.V. All rights reserved PH S 0 0 1 2 - 3 6 5 X ( 9 6 ) 0 0 3 6 6 - 4
224
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243
In the terminology of Wei and Yang [12], we say that a linear [n, k] code C satisfies the chain condition, if there exist r-dimensional subcodes Dr of C for 1 ~
r = 1,2 ..... k,
and D1 C_D2 C • • • C_Dk, Many applications of generalized Hamming weights are known. They are useful in cryptography [2, 11], in trellis coding [7, 13], in truncating a linear block code [5], etc. The generalized Hamming weights have been determined for binary Hamming codes, MDS codes, Golay codes, Reed-Muller codes and their duals [11], Kasami codes [5], the projective codes from nondegenerate quadrics [10], the projective codes from nondegenerate Hermitian varieties [6], the projective codes from degenerate Hermitian varieties [16], and projective Reed-Muller codes of lower orders [15]. Some of these codes are proved to satisfy the chain condition [12]. Let C be a [n, k] linear code. For r, and i, with 1 ~
~""
,A(r).
And the rth generalized weight distribution funtion is the polynomial
A(r)(z) ~- A (r) -4- A~r)z At-'" + A~r)z n. We denote by Ai, 0 <<.i <<.n, the number of codewords in C of weight i. Then the sequence Ao,A1 . . . . . An,
is the weight spectrum of C. Obviously, Ao = 1,
Ai = ( q -
1),411), i = 1..... n.
The support weight distribution for irreducible cyclic codes was introduced by Helleseth et al. [3]. The generalized Hamming weights are actually the minimum support weights and the support weight distribution is exactly the same as the generalized weight spectrum. The weight spectra of the projective codes from nondegenerate quadrics in projective spaces have been determined [10]. Some special cases of the projective codes from degenerate quadrics in projective space have obtained [1]. In the present paper the weight hierarchies and generalized weight spectra of the q-ary projective codes from degenerate quadrics in projective spaces are determined. It is also proved that these codes satisfy the chain condition.
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243
225
2. Preliminaries
Let C be a q-ary linear [n,k] code. The following properties of the generalized Hamming weights are known: (1) (Monotonicity) 1 <<.dl(C) < d2(C) < . . . < dk(C)<~n. (2) (The generalized Singleton bound) dr(C)<~n - k + r for r = 1,2,...,k. (3) (Duality) Let C a- be the dual code of C. Then {dr(Ca-)[ 1 <~r<~n - k} = {1,2 .... , n } \ { n + 1 - d r ( C ) [ 1 <<.r<<.k}.
(4) (The Griesmer-Wei bound) r--1
dr(C)>~--~ [ ~ -i-= 0
/
tJ
/
forr:l,2
. . . . . k,
/
where Ix] denotes the smallest integer not less than x. All these properties are proved for the case q = 2, for (1) cf. [3, 11], for ( 2 ) - ( 4 ) cf. [11]. When q is a power of any prime, the proofs are the same. As in [3] or [8], let G be a generator matrix of C, and for any column vector x E ~:~k), the k-dimensional column vector space over U:q, let mG(x), the number of occurrences of the vector x as columns of G. Obviously, w~(C) = n - m~(0(k)), where 0 (k) is the zero vector of 0:~k). Let U be a subspace of 1:q (k) and define ma(U) = ~
mG(x).
xCU
If M is an r × k matrix of rank r, then M G generate an r-dimensional subcode of C, and any r-dimensional subcode is obtained in this way. Let D be an r-dimensional subcode of C and M G be a generator matrix of D, where M is an r × k matrix of rank r. Define the dual of D to be z) ± =
IMx = 0}.
(1)
Then Da- is an ( k - r)-dimensional subspace. And for any r, where 1 <~r<<.k, the map D ~-* Da- defined by (1) is a bijection from the set of r-dimensional subcodes of C to the set of (k - r)-dimensional subspaces of H:q (k). The following result can be found in [8]. Lenuna 1. L e t D be a subcode o f C. Then ws(D) = n - mG(D ±).
3. The projective code from a degenerate quadric in PG(k -- 1,
IFq)
A linear [n, k] code C is called a projective code if the columns of a generator matrix G of C can be regarded as distinct points of PG(k - 1, UZq), the (k - 1)-dimensional
Z. Wan,X. WulDiscreteMathematics177 (1997)223-243
226
projective space over 0:q. So we have a point set in P G ( k - 1, 0:q), whose elements are the column vectors of G, which will be denoted by Sac,6 and called the point set arising from C via G. Different encoding matrices of C give rise to point sets which are projectively equivalent. Two projective In, k] codes over ~:q are said to be equivalent, if one can be obtained from the other by permuting the coordinates of the codewords and multiplying them by non-zero elements of D:q. Let G be a generator matrix of a projective [n, k] code C and C' be a projective code equivalent to C. Then the same transformation which transforms C to C' will transform the encoding matrix G of C to an encoding matrix G' of C . Clearly, Sac,G = ~c,,G,. It follows that in general, if G and G' are encoding matrices of two equivalent projective [n,k] codes C and C', then Sac,G and Sac,,G, are projectively equivalent. In the sequel, when we consider a fixed code, we denote SaC,G by SaG. Theorem 2. Let C be a q-ary projective [n, k] code, G be a generator matrix o f C, and SaG be the point set in PG(k - 1, ~:q) arising f r o m C via G. For any r, where 1 <~r <<.k, let Dr be an r-dimensional subcode, then there is an (k - r - 1)-flat o f P G ( k - 1, ~:q) Pk-r-1 such that
ws(Dr) = n - IPk_r_l (3 Satl, and dr(C) = n - max{lPk_r_l A SaGI}, where Pk-r-1 runs through all (k - r -
1)-flats o f PG(k - 1, ~:q).
Proof. It follows from Lemma 1. Specializing Lemma of [4] and Lemma 1 of [8] to any q-ary projective [n,k] code C we can also obtain the theorem. [] When q is odd, we choose a fixed nonsquare element z in 0:q, and when q is even, we choose a fixed ~t, where 0~~ {x2+x IxE 0:q}. Let k = 2 v + 3 + l > 0, where ~ = 0, 1, or 2 and assume that v > 0, l>~0. Let Q2v+~+t be the ( 2 v + 6 + l) × ( 2 v + 3 + l) matrix o v e r ~:q,
Q2v+~+t= ( Q2v+~ o(t)), where Q2v+6 is defined as follows. When q is odd,
Q2~=
iOlV I I )
0
'
Q2~+1=
(o,v) I )
0
1
,
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243
and
I(V) 0
I °*) Q2v+2
227
/ 1 --Z
When q is even, Q2v= ( 0 I ~ ) ) ,
0 I (v) 0
Q2v+l =
) 1
I 0~) and
O2v+2 =
i(v) 0
) 1
"
~x
The set of points
t(X1,X 2 . . . .
,X2v+3+l ) satisfying
(X1, X2 . . . . . x2v+3+l)a2v+6+l t(x1, x2 . . . . . X2v+3+l) = 0
is a degenerate quadric in PG(2v + ~ + l - 1, BCq)in its normal form [9], which will also be denoted by Q2v+6+/. Let n = Ia2v+~+tl. Then from [9], we know n=
(q~ - 1)(qV+6-1 +
q-1
qt _ 1 1)q t + - q-1
= q2V+6+l-1 + qv+l _ qV+6+l-1 _ 1
q-1
(,)
For each point of Q2v+6+t, choose a system of coordinates and regard it as a k-dimensional column vector. Arrange these n column vectors in any order into k × n matrix, denote it also by Q2v+6+t. It can be proved that Qzv+6+t is of rank k. Hence Q2v+6+t can be regarded as a generator matrix of a q-ary projective [n, k]-code, which will be denoted by C2v+6+l and called the projective code from the degenerate quadric Q2~,+6+t in P G ( k - 1, U:q). Obviously, ~Q2~,+*+t = a2v+3+l.
4. Flats in P G ( k - 1, IFq) under the singular orthogunal group
Let A and B be two k x k matrices over ~:q. We introduce the notation A=B,
which means that A = B when q is odd, and that A + B is an alternate matrix when q is even. Moreover, A and B are said to be cogredient if there is a k x k nonsingular matrix R such that tRAR = B.
228
Z. Wan, X.. WulDiscrete Mathematics 177 (1997) 223-243
Denote by GL2v+a+t(Brq) the general linear group of degree 2v + 3 + l over DZq.Let
o2v+~+M:q)=
{T [ reGL2v+~+t0:q) and tra2v+~+tr - a2v+~+t}.
Clearly, 02~+~+tis a subgroup o f GL2,+~+z(HZq) and is called the singular orthogonal group of degree 2v + 6 + l over gCq with respect to Q2,+o+z. Define an action of O2~+~+z(DZq) on the (2v + 3 +/)-dimensional column vector space DZq (2v+~+t) as follows: Ozv+6+l(~q) X ~(2v+6+l) ~ -q ( T , t ( x l . . . . . X2v+6+l)) ~
~(2v+6+1) -q T t ( X l . . . . . X2v+~+l).
The action induces an action on PG(k - 1, Dzq), and the set of flats of PG(k - 1, g:q) is subdivided into orbits under 02~+~+t(~q).Let P be an ( m - 1)-flat, then it corresponds to an m-dimenional subspace of U:)k), and there is a k x m matrix whose m columns form a basis of this subspace, if no ambiguity arises, we denote all of them by P. Let P be an m-dimensional subspace of ~:q(k). When q is odd, the m-dimensional subspace P and its corresponding ( m - 1)-flat are said to be type (m,2s + 7,s,F), where ? = 0,1, or 2, F = 1 or z when ? = 1 and F disappears when y = 0 or 2, if tPQ2~+~+zPis cogredient respectively to
2(Ol ) ,0
I
(,o0
o(ra_2s )
'
1
'
o(m -- 2s-- 1 )
i(s) 10) 0
i(s) I ~) 0
or
Z 0 ( m - 2 s - 1)
) 1
. Z o(m-2s-2)
When q is even, the subspace P and its corresponding fiat P are said to be (m,2s + y,s), where y = 0, 1, or 2, if tPQ2v+6+tPis cogredient to
(O2s+~
0(m-2s-~)
)
type
-
Moreover, when 3 = 1, subspaces and flats of type (m, 2s + 1, s) are further subdivided into two types, namely, type (m,2s + 1,s, 1) and type (m,2s + 1,s,0) according to P not containing or containing a vector of the form t(0 . . . . . 0, l , * . . . . . *), 2v
l
respectively. We use the symbol (m,2s + y,s,F), where F = 0 or 1 when y = 1, and F disappears when 7 = 0 or 2, to cover these four cases.
229
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243
Let E be the subspace of oq~:(2v+6+l) generated by e2v+6+l, e2v+6+2, .. ., e2v+6+l , where t ( 0 , . . . , 0 , 1 , 0 , - . . , 0 ) , then dimE : l. An m-dimensional subspace P or an
ej :
j--1
( m - 1)-flat P is said to be type (m,2s+7,s,F,t) if: (1) P is of type (m,2s+7,s,F) and (2) dim(P A E) = t. Denote by Jl(m,2s + 7,s,F;2v + 6 + l) the set of subspaces or fiats of type (m,2s + 7,s,F), and by Jl(m,2s + 7,s,F,t;2v + 6 + l) the set of subspaces or flats ig(2v+6+l) of type (m,2s+7,s,F,t) in .q or P G ( 2 v + 6 + l, Yq), respectively, and let
N(m,2s + y,s,F;2v + 6 + l) : I ~ ( m , 2 s + 7,s,F;2v + 6 + 1)I, N(m,2s + y,s,F,t;2v + 6 + l) = [JC'(m,2s + 7,s,F,t;2v + 6 +
l)l.
The following results will be used in the sequel and can be found in [9]. Lemma 3. Let q be odd. Then Jll(m,2s + 7,s,Y,t; 2v + 6 + l ) is nonempty if and only
if t <<.l and 2s+7<<.m-t~
v + s + m i n { 6 , y } if f ¢ l, o r y # l, ory=f=landF=l, v+s ifT=6=landF=z.
(2)
Moreover, if Jlg(m,2s+ y,s,F,t;2v+6+l) is nonempty, then it is an orbit of subspaces or fiats under 02v+~+l(~:q). Lemma 4. Let q be odd. Then Jg(m,2s + 7,s,F;2v + ~ + l) is nonempty if and
only if min{l,m-2s-7}>>,
{
max{O,m-v-s-min{6,7}} if f # l , o r T # l, (3) or7=6= landF=l, max{O,m- v - s} i f y = 6 : l a n d F = z .
Moreover, ~[(m,2s + 7,s,F;2v + 6 + l) is the disjoint union of all the orbits ~[(m,2s + y,s,F,t;2v + 6 + l) under 02v+~+Z(Fq), where t satisfies (2). Lemma 5. Let q be even. Then ~'(m,2s + 7,s,F,t;2v + 6 + l) is nonempty if and
only if t <~1 and v+s+min{6,7} 2s+7<~m-t<~
if6#l,
ory=f=landF=l, v+s
orT#l, (4)
if y = f = l a n d F = O .
Moreover, if ~q(m,2s+y,s,F,t;2v+6+l) is nonempty, then it is an orbit of subspaees or flats under 02v+a+t(~q).
230
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243
Lenuna 6. Let q be even. Then J / ( m , 2s + ?, s, F; 2v + 6 + l) is nonempty if and only
if
{
max{O,m-v-s-min{6,?}}
m i n { l , m - 2 s - ?} >>.
if 6 # l, o r ? # l,
or v = 3 = 1 and F = 1, max{0,m-v-s} if?=6=landF=0.
(5)
Moreover, ~g(m,2s + %s,F;2v + 6 + l) is the disjoint union o f all the orbits Jg(m,2s + V,s,F,t;2v + 6 + 1) under 02v+6+t(U:q), where t satisfies (4). Now, the numbers o f subspaces or flats are given as follows: I.emma 7. Suppose that t<<.l and that (2) or (4) holds when q is odd or even, respectively, then N(m,2s + y,s,F,t;2v + 6 + l) = N(m - t,2s + y,s,F;2v + 6) l × q(m-t)(l-t) Hi=l-t+l (qi _ 1)
I'Ii=l (qi__ 1)
where N(m - t,2s + %s,F;2v + 6) = q2S(r+s--m+t)+s(6+7)--7(m--t--2s-7) ×
HiVv+s_m+t+T+l (qi __ 1)(qi+6-1 + 1) s - 1 ~' -[--[s+~,--1 $l--[m--t--2s--? (qi__ 1) U i = I (qi lXi=0 ( q i + 1,l/,i___l
× no(m -
t,2s + ?,s,F;2v + 6),
where no(m - t,2s, s;2v + 6) = 1; no(m - t,2s + 1,s,F;2v + 6) qV-S-1 (qV+S-m+t+l __ =
1) if6=o,
qV-S(qV+S-m+t+l _ 1)
if 6 = l and F = z ,
qV-S(qV+S-m+t+l + 1)
if6=
qV-S(qV+S-m+t+2 + 1)
if6=2,
l and F = 1,
when q is odd; and no(m - t,2s + 1,s,F;2v + 6) :-
2 q V - s - l ( q v+s-m+t+l -- 1)
if6=o,
2qm-t-2s-l(q 2(v+s-m+t+l) -- 1)
if6=
2qm-,-Z -I
i f 6 = 1 and F = 1,
2qV-S(q v+s-m+t+2 + 1)
if6=2,
l and F =O,
'
Z. Wan, ~
231
Wu/Discrete Mathematics 177 (1997) 223-243
when q is even; and no(m - t,2s + 2,s; 2v + 6) ~-
{ q2CV-~)-2(qV+S-,n+t+t _ t )(qV+S-m+t+2 _ 1)
if 8 = O,
q2(V-s)-l(qV+S-m+t+2 _ 1)(qV+S-m+t+2 _.}_ 1)
i f 8 = 1,
q2(V-S)(qV+S-m+t+2 q_ 1)(qV+S-m+t+3 q_ 1)
if 8 = 2.
8. Let q be odd. Suppose that (3) holds, then
Lemma
N(m, 2s + ?, s, F; 2v + 8 + l) = ~ N(m, 2s + 7, s, F, t; 2v + 8 + l), t
where the summation range of t takes all values in [max{0, m - v - s - m i n { 8 , 7 } } ,
if8¢
1, or 7 ¢ 1, or 6 = 7 =
min{l,m
-
2s -
7}],
1 and F = 1,
or
[max{O,m - v - s } , if7=8=1
m i n { l , m - 2s - ?}],
and F =z. 9. Let q be even. Suppose that (5) holds, then
Lemma
N(m,2s + 7,s,F;2v + 8 + l) = ~ N(m,2s + ?,s,F,t;2v + 8 + l), t
where the summation ranoe of t takes all values in [max{O, m - v - s - m i n { 8 , V}}, m i n { l , m - 2s - 7}],
if8¢
1, or 7 ¢ 1, or 8 = 7 =
1 and F = 1,
or
[max{O,m - v - s } , if?=8=
5.
Some
Lemma
m i n { l , m - 2s - 7}],
1 andF=O.
lemmas
10. Let m>>.l and P be an ( m -
1)-flat in P G ( 2 v + 6 +
l-
1, UZq). Then
Ie co Q=v+a+tl = I tpQ2~+a+tel • P r o o f . Let x = t ( x l , x 2 . . . . . x2v+6+1 ) be a point o f P G ( 2 v + 6 + l - 1, DZq), and x E P N Q2v+~+z, then x C P and txQ2v+~+zx = 0. F r o m x E P we deduce that x = Py, where y = t ( y l , y2 . . . . . Ym). T h e n tytpQ2v+~+tPy=O, that is y E tpQ2v+~+tP.
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243
232
Conversely, let y -- t(yl, y2 . . . . . Ym) be a point of PG(m - 1, ~:q) and y C tpQ2v+6+lP, then tytpQ2~+6+tPy = 0. Let x = t(xl,x2 . . . . . x2~+6+z) = Py. Then txQ2v+6+lX = O. So, there exist a bijective from tpQ2v+6+tP to P N Q2~+6+t which maps yEtpQ2,+6+tP to x = P y E P f~ Q2v+6+t. [] L e m m a 11. Let m~>l, and PbP2 both be ( m - 1)-flats o f P G ( k - 1,Brq), which belong to the same orbit under 02v+6+l(~:q). Then
I'elQ2v+6+tea[ I'e202 +6+ e21. =
Proof. /'1 and P2 belong to the same orbit under 02v+6+l(~Cq) if and only if there is an m xm nonsingular matrix A and element T E Q2:+6+z(0Zq), such that P1 = TP2A. Thus Itel a2~+a+tP11 = ItA tp2tTQ2v+6+tTP2AI = ItA tp2Q2v+6+lP2AI
= [te2Q2~+a+te2l.
[]
L e m m a 12. Let m>~l and P be an ( m - 1)-flat o f type (m,2s + 7,s,F,t). Then
qm-I + qm-S-r _ qm-S-1 _ 1 IP n Q2v+~+t[ :
q-1
which is independent o f F and t. Proof. By Lemma 10, we have [P N
Q2v+6+tl=
ItpQ2v+6+tPI•
(o1,> /
By L e m m a 11, when q is odd, we can assume that
tpQ2~+6+tp=
I )
0
F
o(m-2s-~) where
F =
0
ifT=0,
(1) or (z)
if Y = 1,
(1-z)
if y = 2 .
(0/, /
And when q is even, we can assume that
tpQ2v+6+lp =
0
F
o(m-2s-r)
'
'
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243
233
where
0
if), = 0 ,
(1)
i f ) , = 1, 1)
if 7 = 2 .
So, we obtain (q~ _ 1)(qS+r-1 + 1)q(m_2s_~ ) + qm-2s-7 _ 1 ItpQ2v+6+IPI =
q-1
q-1
--'__qm-1 + qm-S-7 _ qm-S-1 _
[]
q-1
Corollary 13. Let m>>,l and P be an (m - 1)-flat of type (m,2s + 7,s,F). Then qm-1 I e n Q2,+~+tl =
qm-S-~ +
qm-S-I - 1
q - 1-
which is independent of F.
Corollary 14. Let m>~l and P be an ( m - 1)-flat of type (m,2s+ 1,s,F,t). Then
[P n
qm-1 _ 1 QEv+6+l[ -
q-1
which is independent of s and t.
6. The weight hierarchy of C2,+6+t Lenuna 15. Assume that t <~l, 0 < m < 2v + 6 + l, and that subspaces of type (m, 2s + Then max{lPm_l n Q2,+6+11}, where P~-I runs throu#h all 1,s,F,t) exist in n:(2v+6+t) uq ( m - 1)-flats of P G ( k - 1,~:q), cannot be achieved by ( m - 1)-flats of type (m,2s + 1,s,F,t), unless, 6 -- 2, m = 2v + 1 + l, and t = l. In the later case it is actually achieved by (2v + l)-flat of type (2v + 1 + 1,2v + 1,v,F, l). Proof. Let Pt be an ( m - 1)-flat of type (m,2s+ 1,s,F,t), where t<~l. When q is odd, (2) becomes v + s + min{6, 1} 2s+l<.m-t<
{
if6=0, v+s
or6=2,
if 6 = 1
orf=landF=
andF=z.
1,
(6)
234
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243
When q is even, (4) becomes v + s + min{~5, 1}
2s+l<<.m-t<<.
if ~ = 0, or 3 = 2, or 3 = 1 and F = 1,
v+s
if 3 = 1
(7)
andF=0.
We distinguish the following cases: (a) 6 = 0 , o r 6 - - 1 a n d F = z w h e n q i s o d d , Then both (6) and (7) become
or6=
1 andF=0whenqiseven.
2s+ l~m-t<<.v+s. Let R be one of them. By Thus subspaces of type (m,2s, s,t) exist in 11~(2v+6+1) °q Lemma 12,
qm-1 + qm-s [R n Q2~+6+/1 =
_ qm-S-
1
--
1
q-1
But qm-1 _ 1
I~ n
Q2v+a+tl -
q-1
so IPt n Q2~+~+ll < IR n Q2~+~+zl, hence, max{IPm_l n Q2~+a+zl} cannot be achieved by Pt. (b) 6 -- 2, or 6 = 1 and F = 1. Then both (6) and (7) become 2s+ l~
1.
We distinguish further the following cases: (b.1)m-t
1.
(i) If s < v, then 2 ( s + 1)~
1).
IlZ(2v+6+l) AS in case (a), Hence subspaces of type of (m,2(s + 1),s + 1,t) exist in .q max{IP,~_l n Q2v+6+tl} cannot be achieved by Pt. (ii) I f s = v , t h e n m - t = 2 v + l , i . e . , m = 2 v + l + t , andP~ is of type ( 2 v + l + t,2v + 1, v, F, t). In PG(2v + 6 + l - 1, Fq), all ( m - 1)-fiats which may exist are of type ( 2 v + 1 + t , 2 v + 1,v,F,t) or of type ( 2 v + 1 + t , 2 s ' + v ' , s ' , F , t ' ) , where t'<<.l. Let Wt,
Z. Wan, X.. WulDiscrete Mathematics 177 (1997) 223-243
235
be a subspace o f (2v + 1 + t,2sl+ 7~,s',F,t~). Then by Corollary 14 and Lemma 12,
q2V+t 1
I~ n
Q2v+~+zl -
q-1 q2V+t _[_ q2v+l+t-s'-7 t _ q2V+l+t-s'-I
IW, n Q2v+a+zl
=
_ 1
q-1
When 71/> 1,
IW,, n Q2v+ +tl
IP, n Q2v+ + l.
Because we are studying whether max{IP,n-i n Q2,+~+ll} can be achieved by Pt, Wt, can be neglected. When y/ = 0, Wt, is o f type (2v + 1 + t, 2s 1, s I, t') and q2V+t q_ q2V+l+t-s' _ q2V+t-s' _ 1
IW,, n Q2v++ll
=
q-I
By (2) and (3), Wt, existin PG(2v+ 6 + I- 1,Eq) if and only if 2s/ <.2v + 1 + t - t' <~v + s t, which is equivalent to
v-F 1 + t - t' <~s' <.v + ½(1 + t - t ' ) , from which it follows that
tl>~t + 1. But
IS,
n Q2v+6+ll = q2V+t -t- q2V+l+t-s' _ q2V+t-s' _ 1
q-1 q2V+t + q2V+t-S'(q _
1) --
1
q-1 <~
q2V+t q_ q V - l + t ' ( q _
1) --
1
q-1
therefore when t t = l, [We N Qzv+6+t[ attains its maximum
q2v+l-1 + qV+t-l(q_ 1 ) - 1 [ Wt N az~+a+ll =
q-1
On the other hand, for all t ~
I~ n
q2V+t _ 1 Q2~+I+/I -
-
q-1
q2V+l-1 + q V + t - a ( q _ 1 ) - 1 <
q-1
236
z. Wan. x.. Wu/Discrete Mathematics 177 (1997) 223-243
hence max{lPm_l n O2~+l+Zl} cannot be achieved by Pt. When 6 = 2, we may take t = 1. Then IP! N Q2v+a+/] - q2V+l _ 1 q-1 It is easy to verify that
let n Q2~+a+tl/> IWt n a2~+~+zl. Hence, we conclude that max{IPm_l n a2v+a+zl} is achieved by (2v + / ) - f i a t of type (2v+l+l,2v+l,v,F,l),whenf=2, m=2v+l+l,t=l. [] Lemma 16. Assume that t ~
[]
Lemma 17. Assume that subspaces o f type (m,2s, s,t) and o f type (m,2s',s',t') exist in oq~Z(2v+~+l).Let P and P~ be subspaces o f type (m,2s, s,t) and (m,2sP,s~,t ') respectively. Then IP n Qzv+z+tl > IP'N a2~+6+/I if and only if s < s'. Proof. By Lemma 12,
qm- 1 IP n Q2v+a+tl --
qm--s _ qrn--s- 1 +
- 1
q-1
Clearly, q,,-s _ qm-S- 1 = qm-S- 1(q _ 1), SO the lemma follows.
[]
Theorem 18. The weioht hierarchy o f C2v+a+t is as follows: (a) 6 = 0,
dr(C2v+l) =
q2V-1 _ q2V-r-1 l q-1 q' q2V-1 q2V--r+ q~
r = 1,2,...,v, qV-l ql,
--
q--1
--
r = v + l . . . . . 2v + l,
237
Z, Wan, X. W u l D i s c r e t e M a t h e m a t i c s 177 (1997) 2 2 3 - 2 4 3
(b) 6 = 1,
q2V
q2V-r -
dr(C2v+l+l)
=
q~ + qV-1 ql,
q --1
r = 1 , 2 , . . . , v + 1,
_2v+l--r
q2V
-- tt
1
q-1
q'
r = v + 2 . . . . . 2 v + 1 + l,
(c) 6 = 2,
dl (C2~+2+1) = q~+l(q~ _ 1 ),
q2V+l _
q2V+l-r
_
qa~(C2~+2+l) =
q2V+l
qv+l + qv-1
1
qt,
r=2,3 ..... v+2,
q2V+2-r -- ~v+l ..~ v q-1
r=v+3
q ql,
..... 2v+2+1.
Proof. By Theorem 2,
dr( C2~+a+l ) = n - max{lP2v+6+/_l_r 13 Q2~+6+/1}, where and
e2v+6+l-l-r runs through all (2v + 6 + l - 1 - r)-flats of PG(2v + 6 + l - 1, ~q) q 2 V + 6 + l - 1 ..1_ qV+t _ q V + 6 + l - 1 _ n=
1
q-1
At first, we discuss the case 6 = 2 and r = 1. By Lemmas 15 and 16, we have
dl(C2v+2+t) = n -
[P n
Q2~+6+tl,
where P is ( 2 v + / ) - f l a t o f type ( 2 v + 1 + 1 , 2 v + 1,v,F,l). By Corollary 14, q2v+l _ 1
Ie A
Q2v+a+tl
-
q-1
hence,
q2V+l+t + qV+t _ q~+l+/_ q2V+t dl(C2v+2+l)
-~-
q-1
qV+t(qV :
1). --
Now, we consider all the other cases. By Lemmas 15-17, max{[P2v+6+t-l-r f3 Q2v+6+t[}, where P2~+~+t- l - r runs through all (2v + fi + l - 1 - r)-flats o f PG(2v + 6 + l -- 1, ~q), is achieved by (2v + 6 + l - 1 - r)-fiats o f type (2v + 6 + l - r, 2s, s, t) with s as small as possible. Let P be a ( 2 v + 6 + l 1 - r)-flat of type ( 2 v + 6 + l - r , 2 s , s,t), then
2s<<.2v + 6 + l - r - t<<.v + s. The least possible value o f s satisfying this inequality is v + 6 - r, while t = l, when r<<.v+6, and 0 when r > v + 6 , i.e.,
s={V+6-r, 0,
when r~
v + 6.
238
z. Wan, X. Wu/Discrete Mathematics 177 (1997) 223-243
We distinguish the following cases: (a) 6 = 0. When r<<,v. P is o f type (2v + l -
dr(C2v+l) = q2v+l-I + q~+t _ q~+l-1 _ 1 q-1
r,2(v-
r),v-
r,l). Then
q 2 v + l - r - l + q V + l _ qv+l-I _ 1 q-1
q2V-I _ q2V-r-l ql. q--1 When 2v + l > r > v, P is of type (2v + l - r, 0, 0, l). Then
dr(C2v+l) = q2v+l-I + qV+! _ qv+l-1 _ 1 q--1
q2v+l--r _ 1 q--1
= q2V--1 _ q2V--r + qV _ qV--1 ql. q--1 When r = 2v + l, we have
q2V-1 + qV _ qV-1 d2v+t(C2v+l)-- IQ2v+ll -(b) 6 = 1. When r < ~ v + l ,
- l q l.
q-1
P is o f type ( 2 v + l + l - r , 2 ( v + l
q2V+t _ 1 dr(C2v+l+l) ---- - q--1
-r),v+
1 -r,l),
so
q2v+l-r + q~+I _ qV+t--1 _ 1 q--1
= q2V _ q2V-r _ q~ + qV-1 ql. q--1 When2v+l+l>r>v+l,
P i s of type ( 2 v + l + l - r , 0 , 0 ,
q2V+l _ 1 dr(Czv+l+l) = - q-1
l). Then
q2V+l+l--r _ 1 q-1
q2V __ q2V+ l--r ql" q--1 When r = 2v + 1 + l, we have
d2v+l+l(Czv+l+l) = IQz~+I+Zl- q2V+t _ 1 q-1 (c) 6 = 2. When 2 ~ < r ~ < v + 2 , P is o f type ( 2 v + 2 +
l-r,2(v+2-r),v+2-r,l),
so
q2V+2+l-I + qV+t _ qV+Z+t-1 _ 1 d~(Cz~+z+t) =
q-1 q2V+l _ q2V+l-r _ qV+l + qV-lql. q-1
q2V+l+l-r + qv+l _ qV+l-I _ 1 q-1
Z. Wan, X.. Wu/Discrete Mathematics 177 (1997) 223-243
When2v+2+l>r>v+2,
P i s of type ( 2 v + 2 + l - r , O , O , l ) .
d r ( C 2 v + 2 + l ) = q2V+2+l-I + q ~ + t
_qV+2+l-I
_ 1
q-1 q2V+l _ q2V+2-r __ q V + l + q v
q-1
239
Then
q2V+2+l-r_ 1 q-1
1 •q .
When r = 2v + 2 + l, we have q2V+l+l+qV+l d2v+2+l( f 2v+2+l ) = IO2~+2+zl =
_ qV+l+l _
1
[]
q-1
Corollary 19. Let k = 2v + l, r<<.v. Then dr(C2v+z) meets the Griesmer-Wei bound. Proof. By Theorem 18, dl(Ca~+z)= qZv-2+t. Thus, r--1
Z
[dl(C2v+l)] =qt(q2V-2 + q2V-3 + . . . +q2V-r--1)
i=0 |
q'
/ q2V-1 _ q 2 V - r - l
ql
q-1 = dr(Cav+l).
[]
7. The code C2~+~+t satisfies the chain condition Theorem 20. The code C2~+a+t satisfies the chain condition. Proof. We give only the proof of the case 6 = 2. The proofs of the other two cases are similar. By (2) and (4), there are (2v +/)-fiats of type (2v + 1 + l, 2v + 1, v, F, l) in PG(2v + 1 + l, DZq). Let P2v+z be one of them. Then by Lemma 15,
dl(C2v+2+l) = n - ]P2v+t fq Q2~+2+/1. Again by (2) and (4), there are ( 2 v - 1 +/)-fiats of type (2v + 1,2v, v, l) in PG(2v + 1 + l, DZq). Let P2~-l+z be one of them, and we can assume that P2~-1+1CP2~+l. Then by Lemmas 15-17, d2( C2v+2+l ) = n - ] P 2 v - l + / ( " 1 Q 2 v + 2 + t [ .
And, again by (2) and (4), there are (2v - 2 +/)-flats of type (2v - 1 + 1,2(v - 1), v - 1, l) in PG(2v + 1 + l, 0:q). Let P2~-2+t be one of them, and we can assume that Pz~-z+z CP2v-l+l. Then by Lemmas 15-17,
d3(C2v+2+l) = n - [Pzv-:+l N Q2v+2+l[.
Z. Wan, X. Wu/Discrete Mathematics 177 (1997) 223-243
240
Proceeding in this way, we can find a chain o f fiats P2v+l D P2v-l+l D P2v-2+l D . . . DP0 D P - l , where PEv+t is a (2v + / ) - f l a t of type (2v + 1 + 1,2v + 1, v, F, l), P/ is an/-flat of type (i+ 1,2(i-v+ 1 - l),i-v+ 1 - l , l ) or ( i + 1,0,0, I), when 2 v - 1 + l>~i>~v- 1 + l or v - 1 + I > i/> 0, respectively, and P _ 1 = 0, such that dr(C2v+2+l) = n - IP2~+l+l_r N Q2~+2+tl
for 1 <~r<~2v + 2 + l.
For - 1 <~i<<.2v + l, let Pi± = {t(xl .... ,xk) E D:~qk)l(Xl. . . . . xk) t(yl . . . . . y/,) = 0 V t(yl . . . . . Yk) EP/}, where k = 2v + 2 + l. Then P/± is a (2v + l - / ) - f l a t in PG(2v + 1 + l, 0:q) and ± C P2v-l+l ± P2v+l C .
.C P. ~ .C P~-1 .
1~2v+2+1) q
Denote by P,± a matrix representation of the (2v + l - / ) - f l a t P/± and let C2v+l+l-i = PiZG2v+2+l,
where G2,~+2+t is a generator matrix of the projective code C2~+l+t. Then C2~+l+l-j is a (2v + 1 + l - / ) - d i m e n s i o n a l subcode of C2~+2+t, and we have C1 C C2 C ' " • C C2v+l+l C C2v+2+l.
Then for - 1 <~i <<.2v + l, ]z(f2v+l+l-i)l -- n - IP/fq 02v+2+11 = d2v+l+l-i(C2v+2+l). So Cj, j = 1,2 . . . . . 2v + 2 + / , is a chain of subcodes o f C2v+2+t such that Iz(Cj)I -- dj(C2~+2+l),
1 <~j<~2v + 2 + l,
and Cj c_ Cj+I. Therefore C2~+2+t satisfies the chain condition.
[]
8. The generalized weights spectra of C2~+6+t Theorem 21. For r = 1, 2 . . . . . k - 1, all r-dimensional subcodes o f C2v+a+j, whose duals are o f type ( 2 v + 6 + 1 - r,2s + y,s,F), have the same support weight. Denote such a subcode by Dr:,~,r, then
w,(Or;s,:) q2V+f+l-l+qv+l_}_q2V+6+l-r-s-1 __ qV+6+l-1 _ q2V+6+l-r-I __ q2V+6+l-r-s-~
q-1
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243 Proof. Immediate from Theorem 2 and Corollary 13.
241
[]
For simplicity, write W~;s,O, w~;s,l,r, and Wr;s,2 for ws(Dr;~,o), ws(Dr;s,l,r), and ws(Dr;s,2), respectively. Clearly, Wr;s,l,r does not depend on s and F, and we may write Wr;,1 for Wr;s,l,r. Then Wr;s,O q2V+f+l-lq_qv+ld_q2V+f+l-r-s-1
_ qV+6+l-I _ q2V+f+l-r-I
_ q2V+f+l-r-s
q-1 q2V+f+l-lq_qv+l
_ qV+6+l-1
_ q2V+f+l-r-1
q - 1
Wr;'l ~-
'
Wr; s, 2 q2V+6+l-l_t_qV+l+q2V+6+l-r--s-1
_ qV+6+l-1 __ q 2 V + f + l - r - - I _ q 2 V + f + l - r - s - 2
q-1 Denote by wr the set o f support weights o f r-dimensional subcodes o f C2v+a+l, then Wr is the union of the following three sets {wr;~o,O Is0 satisfies (8)},
(wr,,1}, {w~;s2,2 Is2 satisfies (9)}, where (8) and (9) are min{l, 2v + 6 + l - r - 2so} ~>max{0, v + 6 + l - r - So},
(8)
min{l, 2v + 6 + l - r - 2s2 - 2} >~max{0,v + 6 + l - r - s2},
(9)
and
respectively. Therefore we have
Theorem 22. For r = 1, 2 .... , k - 1, the r-generalized weight distribution function o f C2v+~+t is A(r)(z) = ~ N ( 2 v + 6 + l - r,2so,so;2v + fi + l)zWr:so,° 5o
+ ~ N ( 2 v + 6 + l - r,2sl + 1 , s l , F ; 2 v + 6 + l)z .... . gl,ff
+~N(2v S2
+ 6 + l - r,2s2 + 2,s2;2v + 6 + l)zWr;s2,2,
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243
242
where the summation range of F is {1,z} when q is odd, and is {1,0} when q is even, and the summation ranges of so,s2,sl are given by (8), (9), and m i n { l , 2 v + 6 + l - r - 2sl - 1} max{O, v + 6 + l - r - S 1 tf6¢
-
min{6, 1}},
-
1, o r 6 = l a n d F =
1,
>>. m a x { O , v + 6 + l - r - s l }
(10)
if 6 : 1 and F = z when q is odd, or if 6 = 1 and F : 0 when q is even, respectively. Moreover, A(k)(z) = zWs(G~*+6+,)= z n, where n is given by (*). Proof. We need only to prove that ws(C2v+6+t) = n, but this is trivial.
[]
C o r o n a r y 23. The weight spectrum of C2~+6+t is as follows:
(1) The set of all weights achieved by some nonzero code-word in C2v+6+t is Wo U W1 t_J W2, where
wo r~q2V+6+l-2 _ q2V+6+l-s-2 _ qv+tq 6-~ -- 1 m i n { 1 , 2 v + 6 + l - 1 - 2 s ) q-1 >~max{O,v+6+l 1 -s}f' [ m i n { l , 2 v + 6 + l - 2 - 2s}
W1 = { qZ~+6+t-2 _ q~+tq6-1q-1- 1 />
']
max{O,v+l-l-s},
3=0,1,
max{O,v+l-s},
6=2
,
and
W2 -- {q2V+6+l-2_q2V+6+l-s-3
qV+lq 6 - 1 - 1 q--1
min{l'2v+6+l-3-2s} } >~max{O,v+6+l--l--s} "
(2) For every weight i, denote by Ai the number of codewords which are of weight i.
Then Ao= 1, Ai = O, when i ~ O and i ~ Wo U W1U W2,
Z. Wan, X. WulDiscrete Mathematics 177 (1997) 223-243
243
Ai = ( q - 1 ) ~ N ( 2 v + 6 + l - 1,2s + y , s , F ; 2 v + 6 + l), F,S
when i E Wo U W1 tO W2, where the summation ranges o f s are as in Wo, W1, and W2, F = 1 or z when q is odd, F = 1 or 0 when q is even. []
Acknowledgements The authors would like to thank the referees for their helpful comments.
References [1] I.M. Chakravarti, Families of codes with few distinct weights from singular and non-singular Hermitian varieties and quadrics in projective geometries and Hadamard difference sets and designs associated with two-weight codes, in: IMA vol. 20, Coding Theory and Design Theory (Part I, Coding Theory) 35-50. [2] B. Chor, O. Goldreich, J.Hastad, J. Friedmann, S. Rudish and R. Smolesky, The bit extraction problem of t-resilient functions, Proc. 26th Symp. Foundations of Comput. Sci. (1985) 396-407. [3] T. Helleseth, T. Kl~ve and J. Mykkeltveit, The weight distribution of irreducible cyclic codes with block length nl((q t - 1)/N), Discrete Math. 18 (1977) 179-211. [4] T. Helleseth, T. Klove and O. Ytrehus, Generalized Hamming weights of linear codes, IEEE Trans. Inform. Theory 38 (1992) 1133-1140. [5] T. Helleseth and P.V. Kumar, The weight hierarchy of the Kasami codes, preprint. [6] J.W.P. Hirschfeld, M.A. Tsfasman and S.G. Vladut, The weight hierarchy of higher-dimensional Hermitian codes, IEEE Trans. Inform. Theory IT-40 (1994) 275-278. [7] T. Kasami, T. Takata, T. Fujiwara and Shu Lin, On the optimum bit orders with respect to the state complexity of trellis diagrams for binary linear codes, IEEE Trans. Inform. Theory IT-39 (1993) 242-245. [8] T. Klove, Support weight distribution of linear codes, Discrete Math. 106/107 (1992) 311-316. [9] Z. Wan, Geometry of Classical Groups over Finite Fields (Studentlitteratur, Lund, 1993). [10] Z. Wan, The weight hierarchies of the projective codes from nondegenerate quadrics, Designs, Codes and Cryptography 4 (1994) 283-300. [11] V.K. Wei, Generalized Hamming weight for linear codes, IEEE Trans. Inform. Theory IT-37 (1991) 1412-1418. [12] V.K. Wei and K. Yang, On the generalized Hamming weights of squaring construction and product codes, IEEE Trans. Inform. Theory IT-39 (1993) 1709-1713. [13] J.K. Wolf, Efficient maximum likelihood decoding of linear block codes using a trellis, IEEE Trans. Inform. Theory IT-24 (1978) 76-78. [14] J. Wolfman, Codes projectifs a deux ou trois poids associes aux hyperquadries d'une geometric finie, Discrete Math. 13 (1975) 185-211. [15] X. Wu, Generalized Hamming weights of some classes of projective codes, Ph.D. Dissertation, Institute of Systems Science, Chinese Academy of Sciences, 1995. [16] X. Wu, Generalized Hamming weights of the projective codes from degenerate Hermitian varieties, Science in China (Series E) 2 (1996) 148-158.